Key points from article :
New AI model can diagnose diseases from medical images and explains its reasoning with a visual map.
This transparency allows doctors to confirm the diagnosis and explain it to patients.
Model is trained on various diseases including breast cancer, macular degeneration, and heart enlargement.
Its accuracy is comparable to existing AI models, with success rates above 77%.
The model is inspired by simpler linear neural networks for better interpretability.
Researchers hope to expand its capabilities to diagnose more diseases.
“..our tool benefits society not only in terms of disease diagnoses, but also improving trust and transparency between doctors and patients,” - principal investigator Mark Anastasio.
Research by Beckman Institute at University of Illinois Urbana–Champaign, appeared in IEEE Transactions on Medical Imaging.
